Académique Documents
Professionnel Documents
Culture Documents
The FEM is a novel numerical method used to solve ordinary and partial differential
equations. The method is based on the integration of the terms in the equation to
be solved, in lieu of point discretization schemes like the finite difference method.
The FEM utilizes the method of weighted residuals and integration by parts
(Green-Gauss Theorem) to reduce second order derivatives to first order terms.
The FEM has been used to solve a wide range of problems, and permits physical
domains to be modeled directly using unstructured meshes typically based upon
triangles or quadrilaterals in 2-D and tetrahedrons or hexahedrons in 3-D. The
solution domain is discretized into individual elements – these elements are
operated upon individually and then solved globally using matrix solution
techniques.
The underlying mathematical basis of the finite element method first lies with the
classical Rayleigh-Ritz and variational calculus procedures. These theories
provided the reasons why the finite element method worked well for the class of
problems in which variational statements could be obtained (e.g., linear diffusion
type problems). However,as interest expanded in applying the finite element
method to more types of problems, the use of classical theory to describe such
problems became limited and could not be applied, e.g., fluid-related problems.
Extension of the mathematical basis to non-linear and non-structural problems was
achieved through the method of weighted residuals (MWR), originally conceived
by Galerkin in the early 20th century. The MWR was found to provide the ideal
theoretical basis for a much wider basis of problems as opposed to the Rayleigh-
Ritz method. Basically, the method requires the governing differential equation to
be multiplied by a set of predetermined weights and the resulting product
integrated over space; this integral is required to vanish. Technically, Galerkin's
method is a subset of the general MWR procedure, since various types of weights
can be utilized; in the case of Galerkin's method, the weights are chosen to be the
same as the functions used to define the unknown variables. Most practitioners of
the finite element method now employ Galerkin's method to establish the
approximations to the governing equations.
The versatility of the FEM, along with its rich mathematical formulation and
robustness makes it an ideal numerical method for a wide range of problems. The
ability to model complex geometries using unstructured meshes and employing
elements that can be individually tagged makes the method unique. The ease of
implementing boundary conditions as well as being able to use a wide family of
element types is a definite advantage of the scheme over other methods. In
addition, the FEM can be shown to stem from properly-posed functional
minimization principles.
5. Can The Fem Handle A Wide Range Of Problems, I.e., Solve General Pdes?
While the FEM was initially developed to solve diffusion type problems, i.e., stress-
strain equations or heat conduction, advances over the past several decades have
enabled the FEM to solve advection-dominated problems, including
incompressible as well as compressible fluid flow. Modifications to the basic
procedure (utilizing forms of upwinding for advection, i.e., Petrov-Galerkin and
adaptive meshes) allow general advection-diffusion transport equations to be
accurately solved for a wide range of problems.
6. What Is The Advantage Of The Fem Over Finite Difference (fdm) And Finite
Volume (fvm) Methods?
The major advantages of the FEM over FDM and FVM are its built-in abilities to
handle unstructured meshes, a rich family of element choices, and natural handling
of boundary conditions (especially flux relations). The FDM is generally restricted
to simple geometries in which an orthogonal grid can be constructed; for irregular
geometries, a global transformation of the governing equations (e.g., boundary
fitted coordinates) must be made to create an orthogonal computational domain.
Likewise, implementation of boundary conditions in FDM can be cumbersome. The
FVM is an integral approach (typically with limits -0.5 to 0.5) similar to the FEM,
with volumes being used instead of elements. The divergence theorem is used to
establish the final equation set. Solutions are obtained at volume faces, vertices,
or volume centers – some methods employ staggered grids. While FVM can
handle irregular domains using unstructured grids (stemming from the FEM), the
required averaging over the volume limits the method to second order spatial
accuracy.
7. Is There Any Connection Between The Fem And The Boundary Element Method
(bem)?
In the BEM, one reduces the order of the problem by one, i.e., a two-dimensional
domain is reduced to a line integral – a three-dimensional domain becomes a two-
dimensional surface. The BEM only requires the discretization of the boundaries
of the problem domain – no internal meshing is required, as in the FDM, FVM, and
FEM schemes. The BEM requires two applications of the Green-Gauss Theorem
(versus one in the FEM and employing Galerkin’s Method). The method is ideal
for handling irregular shapes and setting boundaries that may extend to (near)
infinity. One can place interior nodes within the BEM to obtain internal values
easily. The BEM works quite effectively for linear differential equations – principally
elliptic equations. However, if one desires to solve nonlinear advection-diffusion
transport equations, the method becomes very cumbersome and computationally
demanding – BEM matrices are dense, and do not readily permit efficient, sparse
matrix solvers to be used as in the FEM.
Writing a FEM code is not terribly difficult, especially if one develops the code
utilizing a general set of subroutines, e.g., input data, integration, assembly,
boundary conditions,output, etc. About 90% of a FEM program is generic, which is
fairly common among most FEM codes – they tend to use similar matrix solvers,
quadrature rules, and matrix assembly procedures; I/O is usually the major
difference among commercial FEM codes – some are easy, and some are not so
easy to learn and use. A source listing of the FORTRAN codes can be found in the
FORTRAN file folder; flow charts can be obtained from the authors. Likewise,
MATLAB and MathCad files are also available. One of the best commercial
packages now on the market is COMSOL, which also allows users to write their
own solver packages and PDEs.
10. Are There Any Recommended Commercial Fem Packages That Are Versatile In
Handling A Wide Range Of Problems?
Any of the well known and widely versatile FEM codes now on the market
are good – it just depends on how comfortable the user is with the I/O part
of the program. COMSOL,as mentioned before, is quite easy and very
versatile – handling a wide range of problem classes including fluid flow
(with turbulence), heat transfer, structural analysis,electrodynamics, and
general PDEs including species transport, chemical reactions, and
groundwater/porous media flows.
11. How Long Does It Take For Me To Be Able To Use A Fem Program?
Some programs allow you to solve problems fairly quickly. It is always highly
recommended that work out the example problems generally provided by
most commercial software. COMSOL, ANSYS, ALGOR, and NASTRAN all
run on PCs.
The versatility, ease of data input, and solution accuracy make the FEM one
of the best numerical methods for solving engineering problems. FEM
programs are the backbone of structural analyses, and are becoming more
widely accepted for problems in which geometries are complex.
FEM codes range from those that can be found for free on the web to others
costing many thousands of dollars. Those that run on PCs are generally
inexpensive, yet provide powerful tools for solving a number of large scale
problems.
22. What are the factors governing the selection of finite elements?
Nodes and nodal points- The intersection of the different sides of elements
are called nodes. Nodes are of two types - external nodes and internal
nodes.
O External nodes - The nodal point connecting adjacent elements.
O Internal nodes- The extra nodes used to increase the accuracy of the
solution.
Nodal lines - The interface between elements are called nodal
Continuum- The domain in which matter exists at every point is called a
continuum. It can be assumed as having an infinite number of connected
particles.
Primary unknowns- The main unknowns involved in the formulation of the
element properties are known as primary unknowns.
Secondary unknowns- These unknowns are derived from primary
unknowns are known as secondary unknowns. In displacement
formulations, displacements are treated as primary unknowns and stress,
strain, moments and shear force are treated as secondary unknowns.
3-D elements are used for modeling solid bodies and the various 3-
Delements are tetrahedron, hexahedron, and curved rectangular solid.
Element aspect ratio is defined as the ratio of the largest dimension of the
element to its smallest dimension.
Plane strain is a state of strain in which normal strain and shear strain
directed perpendicular to the plane of body is assumed to be zero.
A small unit having definite shape of geometry and nodes is called finite
element.
43. What are the three phases of the finite element method?
45. What are the methods generally associated with the finite element analysis?
The following two methods are generally associated with the finite element
analysis. They are
Force method.
Displacement or stiffness method
47. What is the polynomial type of interpolation functions that are mostly used in FEM?
The polynomial type of interpolation functions are mostly used due to the
following reasons:
It is easy to formulate and computerize the finite element equations.
It is easy to perform differentiation or integration.
The accuracy of the results can be improved by increasing the order of the
polynomial .
Ritz method.
Rayleigh – Ritz method
Name the weighted residual methods.
Point collocation method.
subdomain collocation method.
Least square method
galerkin’s method
Aspect ratio is defined as the ratio of the largest dimension of the element
to the smallest dimension. In many cases, as the aspect ratio increases, the
inaccuracy of the solution increases. The conclusion of many researches is
that the aspect ratio should be close to unity as possible.
The truss elements are part of a truss structure linked together by point
joints, which transmit only axial force to the element.
U=2x2+4y2+6xy
Strain,e= δu/δx= 4x +6y
H version and p versions are used to improve the accuracy of the finite
element method.
In h version, the order of polynomial approximation for all elements is kept
constant and the number of elements is increased.
In p version, the number of elements is maintained constant and the order
of polynomial approximation of element is increased.
58. During discretization, mention the places where it is necessary to place a node
The following places are necessary to place a node during the discretization
process.
Concentrated load-acting point.
Cross section changing point
Different material interjection point
Sudden change in load point.
Static analysis: The solution of the problem does not vary with time is known
as static analysis.
Example: Stress analysis on a beam.
Dynamic analysis: The solution of the problem varies with time is known as
dynamic analysis.
Example: vibration analysis problems.
ANSYS
NASTRAN
COSMOS
NISA
62. Distinguish between potential energy function and potential energy functional.
There are three types of loading acting on the body. They are:
Body force (f)
Traction force (T)
Point load (P)
67. What are the basic steps involved in the finite element modeling.
a. The points in the entire structure are defined using co-ordinate system is
known as global co-ordinate system.
A natural co-ordinate system is used to define any point inside the element
by a set of dimensionless numbers, whose magnitude never exceeds
unity.This system is very useful in assembling of stiffness matrices.
75. How do you calculate the size of the global stiffness matrix?
79. State the assumptions are made while finding the forces in a truss
.
The following assumptions are made while finding the forces in a truss.
All the members are pin jointed.
The truss is loaded only at the joints.
The self-weight of the members is neglected unless stated.
80. Write down the expression of shape function N and displacement u for one-
dimensional bar element.
The total potential energy π of an elastic body, is defined as the sum of total
strain energy U and potential energy of the external forces,(W).
Total potential energy, π = Strain energy (U) + Potential energy of the
external forces (W).
A body is in equilibrium if the internal virtual work equals the external virtual work
for every kinematically admissible displacement field.
Six noded triangular elements are known as linear strain triangle (LST),
which is shown in fig. It has twelve unknown displacement degrees of
freedom. The displacement functions for the element are quadratic instead
of linear as in the CST.
90. Write down the stiffness matrix equation for two-dimensional CST element.
91. Write down the expression for the shape function for a constant strain triangular
element.
95. What are the ways in which a three dimensional problem can be reduced to a two
dimensional approach.
Plane stress: One dimension is too small when compared to the other two
dimensions.
Example: gear – thickness is small
Plane strain: one dimension is too large when compared to the other two
dimensions example : Long pipe [length is long compared to diameter]
Axisymmetric : geometry is symmetric about the axis.
Example: cooling tower
97. Derive the stiffness matrix [K] for the truss element
99. Using finite elements, find the stress distribution in a uniformly tapering bar of
circular cross sectional area 3cm2 and 2 cm2 at their ends, length 100mm,
subjected to an axial tensile load of 50 N at smaller end and fixed at larger end.
Take the value of Young's modulus as 2 x
105N/mm2.
102. Derive the expression for the stiffness matrix for an axisymmetric shell
element.
103. Explain the terms plane stress and plane strain conditions. Give the
constitutive laws for these cases.
104. Derive the element stiffness matrix for a linear isoparametric quadrilateral
element.
If the number of nodes used for defining the geometry is less than the
number of nodes used for defining the displacements known as
isoparametric elements.
If the number of nodes used for defining the geometry is the same as the
number of nodes used for defining the displacements is known as
isoparametric element.